[Solomonov Seminar] 267. Solomonov seminar

[Marko Grobelnik]

V sredo 14. januarja 2015 bo ob 13h Oranzni predavalnici IJS 267. 
Solomonov seminar. Oranzna predavalnica je v drugem nadstropju glavne 
stavbe IJS. Posnetki preteklih seminarjev so na 
http://videolectures.net/solomon/

Tokrat bomo gostili Wraya Buntinea, ki se ukvarja z umetno inteligenco 
in strojnim ucenjem. Wray je studiral v Edinburghu, deloval v San 
Franciscu, kasneje v Helsinkih, v zadnjem obdobju pa je v Avstraliji 
direktor instituta NICTA. Z IJS je sodeloval na vrsti projektov. Tema 
predavanja bojo "Topic Models" za modeliranje besedila in omrezij.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Speaker: Wray Buntine, Monash University, NICTA, Australia

Title:    Experiments with Non-parametric Topic Models

This talk will cover some of our recent work in extended topic models to 
serve as tools in text mining and NLP (and hopefully, later, in IR) when 
some semantic analysis is required.  In some sense our goals are akin to 
the use of Latent Semantic Analysis. The basic theoretical/algorithmic 
tool we have for this is non-parametric Bayesian methods for reasoning 
on hierarchies of probability vectors.
The concepts will be introduced but not the statistical detail. Then 
I'll present some of our KDD 2014 paper (Experiments with Non-parametric 
Topic Models), and some extended work such as "Bibliographic Analysis 
with the Citation Network Topic Model" (ACML 2014) and "Topic 
Segmentation with a Structured Topic Model" (NAACL 2013).  Various 
evaluations and comparisons will be made.  The fully non-parametric 
topic model with burstiness is currently the best performing published 
model by a number of measures and is only a small factor slower in speed 
(and small factor larger in memory) than standard LDA implementations.